Gangireddy Harinatha Reddy
@reddyghr@nbkrist.org
Professor Electronics and Communication Engineering
NBKR Institute of science and technology
RESEARCH, TEACHING, or OTHER INTERESTS
Electrical and Electronic Engineering, Biomedical Engineering, Computer Vision and Pattern Recognition, Artificial Intelligence
Scopus Publications
Scopus Publications
B N Balarami Reddy, G Harinatha Reddy, T Sreenivasula Reddy, Pathipati Srihari, and Bethi Pardhasaradhi
IEEE
Nowadays, autonomous underwater vehicle (AUV) technologies provide localization of AUV s, high-precision 3D measurement mapping, and underwater target tracking. Usually, the AUV consists of various sensors to acquire the dense measurements of the underwater scene to perform target tracking and functionalities. In in-water mobility, the bandwidth is a significant bottleneck, allowing communication with other AUV s and performing centralized target tracking and fusion. Since the communication modules within the AUV are compact, low power, and have low bandwidth, the quantized measurements are transmitted to the fusion center (FC). The sensing devices provide different measurements like range, range rate, azimuth, elevation, and directional cosines corresponding to the scene. Whereas the range measurements are in meters, azimuth measurements range from 0 to 360°• Hence, using a single quantizer with a predefined step size leads to tremendous errors. This paper proposes to deploy an optimal quantizer for every measurement channel and then transmit it to the FC. To explicitly study the quantization effect, we have used linear and optimal quantization techniques which can adaptively choose the levels of the measurements. The extended Kalman filter (EKF) in combination with correlation-free covariance intersection (CI) fusion algorithm is used to attain the global track information. The performance of the proposed method is quantified using the position root mean square error (PRMSE) and compared with the no-quantization state-of-art.
Aswani. Lalitha and G. Harinatha Reddy
IEEE
Wireless Sensor Networks (WSN) have long relied on resource allocation technologies, however most of the existing methods are designed with a single network interface. The Multi-Input Multiple-Output Orthogonal Frequency-Division Multiplexing (MIMO-OFDM) methodology is a relatively new channel estimation method used for varying Quality of Service considerations and for various service types, MIMO employs several parallel channels to transmit these services in a decoupled manner. Constant Bit Rate (CBR) traffic is a high-priority service category that has reduced delay than other types of traffic. Performance indicators such as outage probability, throughput, and access probability of both classes of service are evaluated numerically, and the analysis is offered to determine whether or not the power was allocated appropriately. The channel estimation method is used to analyze the distribution of wireless communication resources. Relay selection, channel assignment, and relay assignment are all used as part of a data-aided estimate procedure to boost overall throughput. The proposed research considers a Quality of Service aided power aware Resource Allocation Optimization Model with Channel Estimation (QoS-PARAO-CE) model for accurate resource allocation with channel estimation to enhance the QoS in WSNs. The proposed model when contrasted with the traditional models performance better.
Vivaksha Mohalia, Pathipati Srihari, Sreenivasula Reddy, G Harinatha Reddy, and Bethi Pardhasaradhi
IEEE
The high-speed Kalman filter (KF) algorithms are essential for robotics, autonomous vehicle, target tracking, and other applications. The dimensions of the state vector and traditional matrix multiplication (complexity of order $\\mathcal{O}\\left(n^{3}\\right)$) are the two main reasons for the computational time of the $K F$ algorithm. Hence, a matrix multiplication accelerator module is needed to accelerate the KF algorithm for higher dimensions of the state vector. In this paper, modified Strassen matrix multiplication (complexity of order $\\mathcal{O}\\left(n^{2.80}\\right)$) is utilized to increase the computational efficiency of the KF algorithm. The number of cycles is evaluated against the dimensions of the KF algorithm to illustrate the proposed methodology. After that, 2D-KF and 3D-KF algorithms targeted on DSP processor TMS320C6678 using C language to ensure real-time processing. The $3 D-K F$ with a state of nine consumes $19.962 \\mathrm{~ms}, 30.47 \\mathrm{~ms}$, and $40.04 \\mathrm{~ms}$ of time by employing the hybrid Strassen, Strassen, and conventional matrix multiplication. It is observed that the usage of hybrid Strassen takes only half of the time provided by conventional multiplications.
Aswani. Lalitha and G. Reddy
The Intelligent Networks and Systems Society
: Blind Source Separation (BSS) is a process of separating a set of source signals from mixed-signal without the help of information of source signals. In some noisy acoustic surroundings, instinctive class detection is completely dependent on vocalization which remains a stimulating task. To identify the definite classes easily, the source signals have to be detached from the mixed signals and this separation procedure is considered as a substantial pre-processing phase before the detection procedure takes place. This research mainly focuses on the issues of BSS in bio-acoustic mixed signals. Independent Component Analysis (ICA) is current technique in the area of BSS that can discrete the mixed-signal and also utilizes Negentropy as its objective function. However, this method is penetrating to the separation matrix and it cannot diverge. So, the bootstrap ICA procedures with Fast and Robust Bootstrap (FRB) method is developed which is applicable for all the signals. The quality of separated source signals using Enhanced-ICA and other algorithms are compared and evaluated according to MATLAB toolbox metrics. The results show that Enhanced-ICA with negentropy is used for finding a maximum non-gaussianity which achieves the BER performances of 0.00019 which is better than existing Discrete Wavelet Transform based BSS (DWT-BSS) and Modified Newton with Improved Animal Migration Optimization (MN-IAMO).
Gurukumar Lokku, G Harinatha Reddy, and M N Giri Prasad
Oxford University Press (OUP)
Abstract Facial expression recognition (FER) is the process of identifying human expressions. People vary in their accuracy at recognizing the emotions of others. Use of technology to help people with emotion recognition is a relatively important research area. Various works have been conducted on automating the recognition of facial expressions. The main intent of this paper is to plan for the FER model with the aid of intelligent techniques. The proposed models consist of steps like data collection, face detection, optimized feature extraction and emotion recognition. Initially, the standard benchmark facial emotion dataset is collected, and it is subjected to face detection. The optimized scale-invariant feature transform (OSIFT) is adopted for feature extraction, in which the key points that are giving unique information are optimized by the hybrid meta-heuristic algorithm. Two meta-heuristic algorithms like spotted hyena optimization and beetle swarm optimization (BSO) are merged to form the proposed spotted hyena-based BSO (SH-BSO). Also, the local tri-directional pattern is extracted, which is further combined with optimized SIFT. Here, the proposed SH-BSO is utilized for optimizing the number of hidden neurons of both deep neural network and convolutional neural network in such a way that the recognition accuracy could attain maximum.
Gurukumar Lokku, G. Harinatha Reddy, and M.N. Giri Prasad
Elsevier BV
Aswani Lalitha and Gangireddy Harinatha Reddy
International Information and Engineering Technology Association
A Wireless Sensor Network (WSN) differs from conventional wireless or wired networks in that it interacts with the environment. Orthogonal Frequency Division Multiplexing (OFDM) was investigated as a possible interface technology for making effective use of bandwidth. Such networks have been proposed for a variety of purposes such as search and rescue, disaster assistance, and smart positioning systems. These applications often require a large number of wireless sensors that are powered by batteries and are designed for long-term, human-free deployment. Collisions between network nodes can significantly degrade performance in WSNs. Although increased bandwidth facilitates wireless access to high data frequencies, it is prohibitively expensive to increase due to spectrum limits. This necessitates making good use of the available bandwidth. OFDM has been considered as a possible interface mechanism for efficiently utilising bandwidth. While many signals available in WSN technology can be employed to mitigate collisions, multi-signal allocations may have a significant impact on the efficiency of multistage communications. Real-time multimedia flow raises the chance of sensor network failures and congestion, which reduces the efficiency of Quality of Service (QoS). The main goal of the Signal Allocation Scheme is to allocate an appropriate number of signals to any node in order to use professional bandwidth and assure QoS. Load balancing is intended to measure and prevent collisions caused by the number of available slots in the frame. Preparation is another important component in preventing collisions because it decreases delay and optimises energy utilisation. In this paper, an Integrated Signal Allocation Model with Effective Collision Resolution Model (ICAM-ECR) is used to deploy non-overlapping signals dynamically for varying application loads based on expected bandwidth estimation. The suggested model is compared to standard methods, and the findings reveal that the proposed model outperforms existing models.
Gurukumar Lokku, G. Harinatha Reddy, and M. N. Giri Prasad
IEEE
Exertion of Transfer Learning for Convolutional Neural Networks (CNNs) describes a dynamic solution for face recognition instead of building and training a neutral network from scratch. Transfer Learning gains and stores the knowledge while solving one problem, and utilizes it for a different related problem. So, choosing a suitable CNN for Deep Face Recognition among the state of art CNN is a challenging task. Deep face recognition requires a large density of datasets with standard subjects. Subjects with parameters like pose, illumination and expression are major constraints for frontal facial images. Ignoring the parameters in training as well as testing phase will result to significant decrease in recognition rate; also show indelible impact on accuracy. Training the network with improper illumination facial data base can lead to biased classifier in deep CNN. In this paper, we propose AlexNet-CNN architecture-based transfer learning framework to enlarge the facial feature space on the constrained face dataset. The proposed framework is self-possessed with input layer, convolutional layers, activation layers, pooling layers, fully connected layer connected to a classifier layer and runs on standard CUHK (China University of Hong Kong) and ORL (Olivetti Research Laboratory) database in a standard frontal pose faces, under ordinary illumination state and with unbiassed expression ensuing to excellent accuracy in facial recognizing tasks.
Face Recognition Technique (FRT) was a unique Biometric Technique which tries to spot the people from still images or video frames by using techniques of pattern recognition. Face recognition includes both face identification and face verification (authentication). The FR Design system follows two basic steps i.e. Feature extraction and classification of patterns. Automated FR finds many applications in real time environment ranging from Social Media, surveillance to biometric authentications. Many state-of-the-art face recognition techniques had been implemented, but the Automated Face Recognition (AFR) taken by digital cameras in unconstraint real‐world environment continues to be terribly difficult, since it involves vital variations in each acquisition conditions, yet as in facial expressions and in pose variations. Thus, this paper presents the theme of computer based automatic face recognition in lightweight of the most contests therein areas with developed solutions that supports applications of signal, image processing and computing strategies.